Experimental Neurology 221 (2010) 98–106

Contents lists available at ScienceDirect

Experimental Neurology

j ourna l homepage: www.e lsev ie r.com/ locate /yexnr

Metallothionein induces a regenerative reactive astrocyte phenotype via JAK/STATand RhoA signalling pathways

Y.K.J. Leung a,1, M. Pankhurst a, S.A. Dunlop b, S. Ray a, J. Dittmann a, E.D. Eaton a, P. Palumaa c, R. Sillard c,M.I. Chuah a, A.K. West a, R.S. Chung a,⁎,1

a Menzies Research Institute, University of Tasmania, Private Bag 58, Hobart, Tasmania 7001, Australiab School of Animal Biology, University of Western Australia, Nedlands, Western Australia 6907, Australiac Department of Gene Technology, Tallinn Technical University. Akadeemia tee 15, Tallinn 12618, Estonia

⁎ Corresponding author. Fax: +61 3 62262703.E-mail address: rschung@utas.edu.au (R.S. Chung).

1 These authors contributed equally to the laboratory

0014-4886/$ – see front matter. Crown Copyright © 20doi:10.1016/j.expneurol.2009.10.006

a b s t r a c t

a r t i c l e i n f o

Article history:Received 7 May 2009Revised 5 October 2009Accepted 5 October 2009Available online 15 October 2009

Keywords:Traumatic brain injuryAstrogliosisRegeneration

Following central nervous system injury, astrocytes rapidly respond by undergoing a stereotypical pattern ofmolecular and morphological alterations termed “reactive” astrogliosis. We have reported previously thatmetallothioneins (MTs) are rapidly expressed by reactive astrocytes and that their secretion and subsequentinteraction with injured neurons leads to improved neuroregeneration. We now demonstrate thatexogenous MT induces a reactive morphology and elevated GFAP expression in cultured astrocytes.Furthermore, these astrogliotic hallmarks were mediated via JAK/STAT and RhoA signalling pathways.However, rather than being inhibitory, MT induced a form of astrogliosis that was permissive to neuriteoutgrowth and which was associated with decreased chondroitin sulphate proteoglycan (CSPG) expression.The results suggest that MT has an important role in mediating permissive astrocytic responses to traumaticbrain injury.

Crown Copyright © 2009 Published by Elsevier Inc. All rights reserved.

Introduction

The most abundantly expressed metallothionein (MT) isoforms inthe brain are MT-I/-II (“MT”). MT is primarily expressed in astrocyteswithin the brain (West et al., 2004), and appears to have importantprotective roles following brain injury. This is demonstrated in studiesutilising experimental approaches to induce differential distributionof MT within the murine brain. In particular, endogenous MT appearsto be involved in regulating astrogliosis following brain injury. Hence,MT overexpressing mice show greater reactive astrogliosis following6-aminonicotinamide (6-AN) administration or cryolesion-inducedcortical injury (Penkowa et al., 2005) compared to normal wildtypes(Penkowa et al., 2002; Giralt et al., 2002). Conversely, MT-I/-IIknockout (null mutant; MTKO) mice exhibit reduced astrogliosisfollowing cryogenic injury (Giralt et al., 2002). The role of MT inmodulating astrogliosis is complex however. This is exemplified bythe finding that while astrogliosis is initially reduced in MTKO mice,reactive astrocytes do develop at later times post-injury than inwildtype mice, and are present for many days longer (Penkowa et al.,1999). This observation of delayed astroglial activation in MTKO micepost-injury has also been observed in a recent microarray study,where GFAP expressionwas noted to bemuch higher inMTKOmice at4 days post-injury than in wildtype mice (Penkowa et al., 2006).

work performed in this study.

09 Published by Elsevier Inc. All ri

Interestingly, intraperitoneal injection of exogenous MT following 6-AN administration or cortical cryolesion greatly enhances astrogliosis(Penkowa et al., 2002; Giralt et al., 2002), suggesting that MT maymodulate astrogliosis by acting as an extra-cellular signallingmolecule.

Following CNS injury, reactive astrogliosis has generally beenconsidered a major impediment to axon regeneration (Bush et al.,1999) due to glial scar formation and the expression of inhibitorymolecules such as chondroitin sulphate proteoglycans (CSPGs) (Yiuand He, 2006). Paradoxically, while the presence of MT appears toenhance astrogliosis, there appears to be a strong correlation betweenMT expression and positive outcomes following CNS injury. Miceoverexpressing MT exhibit improved recovery in a variety of traumamodels (Giralt et al., 2002; Penkowa et al., 2002; van LookerenCampagne et al., 1999). By contrast, in MT-I/-II null mice, cryolesionsincrease neuronal apoptosis and enhance microglial/macrophageresponses within the injury site (Penkowa et al., 1999); similarly, pooroutcomes are seen following kainic acid lesions (Carrasco et al., 2000)and focal ischaemia (Trendelenburg et al., 2002).

MT's beneficial roles in recovery from injury might conceivably beplayed out intracellularly. Indeed, MT proteins lack conventionalsecretion signal peptides and their free radical scavenging and heavymetal binding properties, particularly for zinc, are well characterised.However, other data suggest an extracellular mechanism. We haverecently reported that cultured astrocytes induced into a reactivephenotype by IL-1α and zinc specifically secrete MT (Chung et al.,2008); furthermore exogenous MT acts directly upon injured neurons

ghts reserved.

99Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

to promote axon regeneration in a range of neurotrauma models(Chung et al., 2003, 2008). However, given the accumulating evidencethat MT is involved in regulating astrogliosis, an as yet untestedhypothesis is that secreted MT might also act to promote axonregeneration following injury by modulating astrogliosis.

Here we have examined further whether MT regulates astrogliosisvia extra-cellular mechanisms. We report that exogenous MT doesindeed induce astrogliosis in cultured astrocytes. Furthermore, MTinduces astrogliosis via a JAK/STAT signalling pathway and not theMAPK pathway commonly associated with activation of an inhibitoryastrocytic phenotype. In addition, MT-activated astrocytes showdecreased CSPG expression and promote, rather than inhibit, neuriteoutgrowth in vitro. The data suggest that MT canmodulate astrogliosissuch that reactive astrocytes become permissive to axon regeneration.

Materials and methods

MT null mutant and wildtype mice

MT null mice containing a double deletion of both MT-I and MT-IIgenes were used in this study (Michalska and Choo, 1993). Thesemicehave been maintained on a homozygous C57Bl6 background, soC57Bl6 mice of the same age from a colony housed under the sameconditions were used as wildtype controls.

Primary astrocyte cultures

We obtained highly purified cultures of cortical astrocytes (N95%)using previously established protocols (Chung et al., 2004). Briefly,cerebral cortices dissected from postnatal days 1–3 Hooded Wistarrats or mice (MT null mutant or C57Bl6 wildtype) were transferred toHBSS medium (Sigma). An equal volume of trypsin (0.25% finalconcentration; Sigma) was added, and the cell suspension incubatedat 37 °C for 25 min. Medium was replaced with 2 ml DMEM+10%foetal calf serum (DMEM-10S; both from Invitrogen) and tissuetriturated. The cell suspension was filtered through a 0.6 mm gauzefilter to remove any undigested tissue; the solution made up to 10 mlwith DMEM-10S and centrifuged for 10 min at 500g at 4 °C. Thesupernatant was removed and 1 ml of fresh DMEM-10S added,followed by gentle trituration. This cell suspension was added to apoly-L-lysine (Sigma) coated (1:25 dilution) 75 cm2

flask containing9 ml of pre-warmed DMEM-10S. Cells were incubated for 24 h in 5%CO2 at 37 °C before the media was replaced. Media was then replacedevery 48 h.

Once the cells had become confluent (approximately 8–10 days),the flask was shaken at 250 rpm, 37 °C for 24 h. The medium wasreplaced with 10 ml of fresh DMEM-10S containing 100 ml of pre-warmed Ara-C (Sigma). This was repeated for a further 48 h before thecells were split into two 75 cm2

flasks. Flasks were incubated as beforeand the medium replaced after 24 h, then every 48 h and whenconfluent the astrocytes were plated into either 75 cm2

flasks or 12well tissue culture plates for subsequent experiments. Prior toexperimentation, the confluent cultures were maintained in serum-free medium for at least 3 days.

Primary cortical neuron cultures

Cortical neuron cultures of high purity (N95%) were prepared asreported previously (Chung et al., 2002a,b). Briefly, cortical tissue wasremoved from embryonic day 17 rats, homogenised and trypsinised(0.01%) for 20 min. Following three washes with 10 mM HEPES toremove the trypsin, cells were plated directly onto confluent astrocytecultures (see above) at a density of 1×104 cells/well. The culturemedium consisted of Neurobasal™ medium (Invitrogen), supple-mentedwith 0.1% (f/c) B-27 supplement (Invitrogen), 0.1 mM (f/c) L-glutamine (Invitrogen), and 200 U/ml gentamicin (Invitrogen).

Metallothionein preparation

Rabbit MT-IIA (the most highly expressed isoform of MT in thebrain) was provided by Bestenbalt LLC (Estonia) as a N98% pure HPLCpurified apo-thionein. The MT-IIA was reconstituted with 7.5mol zincsulphate (hereafter termed “MT”) and stored at –20 °C until required.MT addition to cultured cells was performed in the range 1–10 μg/ml,which we (Chung et al., 2003, 2008) and others (Kohler et al., 2003)have shown to promote regenerative neuronal growth.

RNA extraction and reverse transcription

RNA was extracted from astrocytes using Tri Reagent (Sigma),following the manufacturer's instructions. Isolated RNA was quanti-fied by UV spectrophotometry, and RNA quality determined by A260/A280 ratio. cDNA was prepared using Superscript III ReverseTranscriptase enzyme (Invitrogen), and the cDNA stored at –20 °Cuntil required.

Real-time PCR analysis

Real-time PCR primers were designed using the Primer3 programavailable at the following website (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) and ordered from Sigma. The followingPCR primer pairs were used:

Rat GAPDH forward-5′TGCCACTCAGAAGACTGTGG 3′Rat GAPDH reverse-5′GGATGCAGGGATGATGTTCT 3′Rat GFAP forward-5′GGTGGAGAGGACAATCTCA 3′Rat GFAP reverse-5′CCTCTTGTTGGACCGACACA 3′Rat MT-I forward-5′ACCTCCTGCAAGAAGAGCTG 3′Rat MT-I reverse-5′TCACTGTTGTCACGACGAC 3′Rat MT-II forward-5′CACAGATGGATCCTGCTCCT 3′Rat MT-II reverse-5′TGGAGGACGTTCTTTTCGAC 3′

PCR reactions were performed using whole rat brain cDNA toconfirm that only a single band was produced at the expected size foreach PCR primer pair. These PCR products were extracted from agarosegels, quantified by UV spectrophotometry and increasing amounts ofcDNAwere used in real-time PCR reactions to generate standard curvesfor each PCR primer set. cDNA samples were used in real-time PCRreactions using SYBR Green reagent (Ambion) and run on a RotorGenePCRmachine (Corbett Research). Duplicate reactions were run for eachcDNA sample and the ct values for each real-time PCR reaction werecompared to the pre-generated standard curves to determine total copynumber. The total copy number of the target gene was divided by thetotal copy number of GAPDH transcripts to provide a measure of theexpression of the target gene relative to GAPDH expression. Fourdifferent astrocyte cultures were used in four experimental replicates.

Protein extraction, SDS-PAGE and western blotting

Astrocytes were resuspended in cell lysis buffer (20 mM Tris, pH7.4, 200 mM NaCl, 7.5 mM MgCl2, 0.2 mM EDTA, 0.75% Ipegal, 10%glycerol), gently triturated and stored at –20 °C. The proteinconcentration of each sample was determined using the Bradfordprotein assay, and equal amounts of protein loaded for each sampleonto NuPage 10% Bis-Tris pre-cast SDS-PAGE gels (Invitrogen).Western blotting was performed as reported previously (Chung etal., 2002a). Briefly, samples were mixed with LDS loading buffer(Invitrogen) and incubated at 70 °C for 10min prior to loading. The gelwas run at 200V for approximately 25min, removed from the pre-castcasing and proteins electrotransferred to a nitrocellulose membraneusing a western transfer module. Transfer was performed overnight at4 °C at 20V. The membrane was then washed three times in 0.01% PBS

100 Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

with 0.1% tween 20 (PBS-T) with shaking at room temperature.Following blocking in 5% skim milk powder (in PBS-T) for 1 h, themembrane was again washed three times in PBS-T followed by a 1 hincubationwith the primary antibodies (mouse anti-MT 1:500, DAKO;rabbit anti-GFAP 1:500, DAKO; mouse anti-neurocan 1:1000, Chemi-con; mouse anti-phosphacan 1:1000, Chemicon; or rabbit anti-bactin1:5000, Chemicon) diluted in PBS-T. Following a further three washesin PBS-T, the membrane was incubated with the secondary antibodies(1:1000; goat anti-mouse HRP and goat anti-rabbit HRP, Chemicon) in2.5% skim milk powder (diluted in PBS-T) for 1 h shaking at roomtemperature. Following a final three washes in PBS-T, the membranewas incubated with Supersignal™ West Pico Chemi-luminescentSubstrate (Pierce) for 5 min, and then exposed to ECL film fordetection. Four different astrocyte cultures were used to provide fourexperimental replicates.

RhoA activation assay

To measure levels of activated RhoA (GTP-RhoA), at the appropri-ate time post MT-treatment, astrocytes were washed twice in coldPBS, and lysis buffer (125 mM HEPES, pH7.5, 750 mM NaCl, 5% NP-40,50mMMgCl2, 5mMEDTA, and 10% glycerol) added to the cells, whichwere then kept on ice for 20 min. The cell lysate was collected andcleared by centrifugation (10 min, 14,000×g at 4 °C). Half of thevolume of the cell extracts (the other half reserved for total RhoAdetection) was incubated for 1 h at 4 °C with GST-tagged rhotekinRho-binding domain (RBD) bound to glutathione-agarose beads,(MBL). After three washes in lysis buffer, beads were resuspended in2× SDS-PAGE loading buffer and boiled, then beads and supernatantwere loaded onto 10% Nu-PAGE Tris-glycine precast gels (Invitrogen)and subjected to SDS-PAGE and Western blotting (as describedabove) for RhoA. As a positive control, levels of total RhoA weredetermined from the reserved cell lysates not exposed to rhotekin byWestern blotting.

Immunocytochemistry of cultured cells

At the appropriate time, cells were fixed with 4% paraformalde-hyde for 20min. Coverslips were incubated for 1 hwith a combinationof mouse/rabbit primary antibodies diluted in 0.1% PBS, 0.03%tritonX-100™. Antibodies used were anti-GFAP (1:500 dilution;rabbit polyclonal; DAKO) and anti-MT (1:500; mouse monoclonal;Dako). Coverslips were extensively washed in PBS and incubated for1 h with two secondary antibodies (1:1,000 dilution; horse anti-mouse IgG conjugated to Alexafluor 488 and goat anti-rabbit IgGconjugated to Alexafluor 594; Molecular Probes, Eugene, OR), appliedin PBS. Coverslips were mounted onto slides using Permafluormounting medium (Immunotech, Marseilles, France).

Neurite outgrowth analysis

Neurite outgrowth analysis was performed as reported previously(Chung et al., 2007). Briefly, cultures were fixed with 4% paraformal-dehyde and immunolabelled for the neuron-specific cytoskeletalprotein βIII-tubulin and detected using an AlexaFluor-488 secondaryantibody. Digital images were captured on an Olympus BL-50 uprightmicroscope using a Magnafire CCD camera. The images were importedinto HCA-Vision (CSIRO Australia), and the following parameters usedfor the program to automatically identify andmeasure neurites: neuronbody channel=1, size of prefilter=6, top hat width=99, minimumarea of neuron body=225, minimum radius of neuron body=7.

Quantitation of GFAP and CSPG expression by immunocytochemistry

Astrocytes were plated into 24 well plates at a density of 1×104

cells/well in DMEM-10S. After the cells became confluent (3 to

4 days), serum-free medium was applied to the cultures for threedays, after which cells were treated with either MT or an equivalentvolume of vehicle (saline). In preliminary experiments we did notobserve any difference in GFAP or CSPG expression betweenuntreated and saline treated cells (results not shown), and hence insubsequent experiments only vehicle controls were performed.Twenty-four hours after MT treatment, the cells were fixed with 4%paraformaldehyde. Cells were washed with PBS-0.1% triton-X, andendogenous peroxidases quenched by incubation with 0.3% hydrogenperoxide. Blocking solution (1% BSA in PBS) was added for 30 min,followed by addition of the polyclonal rabbit anti-GFAP (1:1000;DAKO) or polyclonal rabbit anti-CSPG (1:500; Santa Cruz) andsecondary anti-rabbit HRP-conjugated (1:2000) antibodies, both for1 h. HRP was detected using the TMB-detection system (KPL),resulting in a blue product. 50 μl of the product was transferred to a96-well plate for analysis by plate reader.

Statistical analysis

For neurite outgrowth analysis, the average total neurite out-growth of each cell was calculated from more than 200 neurons pertreatment for each experiment. Statistical analysis was determinedusing multivariate ANOVA tests (SPSS 16.0), with pb0.05 consideredstatistically significant. For immunocytochemical quantitation, at leastfour replicates of each experimental condition were performed, andstatistical significance determined by ANOVA as described above.

Results

Astrocytic uptake of exogenous MT from culture medium

We firstly investigated the interaction of exogenous MT withcultured astrocytes. Rat cortical astrocytes were maintained in serum-free medium for 4 days, at which time 10 μg/ml MT tagged with thefluorescent dye Alexa-Fluor 594 (MT594) was added to the culturemedium. After 6 h, MT594was observedwithin all astrocytes in a verydistinctive distribution (Figs. 1A, B). MT594 exhibited a punctateperinuclear localisation surrounding, but not inside, the nucleus, andforming a crescent-shaped halo that did not extend throughout thecell body (Fig. 1B). Addition of Alexa-Fluor 594 dye alone did notresult in astrocytic uptake, up to 24 h (Fig. 1C). To distinguishendogenous MT expression and further confirm MT uptake byastrocytes, we added 10 μg/ml non-labelledMT to cultured astrocytesderived from MT-I/-II null mutant mice (Michalska and Choo, 1993)and following immunocytochemical labelling with an anti-MT-antibody we observed a granular distribution of MT within thecytoplasm after 6 h (Figs. 1D, E). While this perinuclear localisationwas similar to the MT-594 experiments, MT immunolabelling wasmuch more granular than MT594 distribution. This may reflectdifferences between live cell imaging (MT594) and the permeabilisingimmunocytochemical techniques. It is also possible that the fluor-ophore labelling has altered the binding and uptake properties of MT.Cytoplasmic extracts were also prepared from MT-I/-II null mutantastrocytes following MT treatment, demonstrating cytoplasmicaccumulation (Fig. 1F). Taken together, these short-term experimentsshowed that exogenous MT is taken up by astrocytes in vitro.

MT treatment induces morphological changes and upregulation ofGFAP and MT-I/-II in cultured astrocytes

Rat astrocytes were grown to confluence, and maintained in serum-free medium for 4 days whereupon they appeared as large polyglonal-shaped flat cells with few processes (Fig. 2A). Upon addition of 5 μg/mlof MT to the culture medium for 24 h, the intensity of GFAPimmunostainingwas significantly increased and astrocytic morphologywas altered now being ramified with numerous fine and spidery, but

Fig. 1. Astrocytes internalise MT594 added to the culture medium within 6 h (A). At higher magnification (B), MT594 was observed in a punctate distribution surrounding thenucleus, which did not extend throughout the cell body. AlexaFluor 594 dye alonewas not internalised by astrocytes after 24 h (C). In a separate series of experiments, unlabelledMTwas added to cultured astrocytes from MT-I/-II null mutant mice (to overcome issues associated with endogenous MT expression), followed by detection of MT byimmunocytochemical techniques. While a faint level of non-specific MT staining was observed in untreated astrocytes from MT-I/-II null mutant mice (D), MT was observed in apunctate distribution across the cell body 24 h after MT treatment (E). Cytoplasmic extracts were collected from MT-treated astrocytes fromMT-I/-II null mutant mice and westernblotting demonstrated the presence of MT associated with these cells at four and 24 h after exposure to MT (F). 10 μg total protein per sample was loaded for the western blottingexperiments. Scale bars=25 μm (A), 10 μm (B–E).

101Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

distinct, GFAP-positive processes emanating from the cell body (Fig.2B); both increased GFAP expression and ramified morphology arecharacteristic of astrogliosis.Western blotting confirmed elevated levelsof GFAP expression following treatment with MT (Fig. 2C).

Exogenous MTwas also found to induce expression of endogenousMT-I/-II by rat astrocytes. Using real-time PCR, it was observed thatMT treatment induced substantial upregulation of MT-I and MT–II

Fig. 2. Confluent astrocytes stained for GFAP displayed a large polygonal-shaped morphologyGFAP immunostaining was significantly increased and astrocytes altered morphology becomcell body (B). Note that images A and B were captured at the same camera intensity settinglevels of GFAPwithin astrocytes after MT treatment (C). 10 μg total protein per sample was loblotting for β-actin. Scale bars=20 μm (A, B).

genes (Figs. 3A, B) 24 h post-treatment. To control for the possibilitythat zinc bound to the MT might be responsible for inducing MTexpression, we performed treatments with an equivalent amount ofzinc (1 nM) that would be bound to the exogenousMT, and found thatthis did not alter the transcript level of either MT genes (Figs. 3A, B),indicating that this upregulation was a specific response to MTtreatment. Immunocytochemistry (Figs. 3C, D) and western blotting

with few processes (A). Following treatment with 5 μg/ml MT for 24 h, the intensity ofing ramified with numerous fine, spidery GFAP-positive processes emanating from the. Western blotting (representative of four different experiments) confirmed increasedaded for the western blotting experiments, and equal loading was confirmed bywestern

Fig. 3. Confluent astrocytes were treated with 1 μg/mlMT, and endogenous MT-I andMT-II expression levels determined by real-time PCR. MT treatment induced expression of bothMT-I (A) and MT-II (B) genes, however an equivalent amount of zinc (1 nM) did not alter gene expression (A, B). Immunocytochemistry (C, D) and western blotting (E)demonstrated elevated levels of MT within astrocytes. 10 μg total protein per sample was loaded for the western blotting experiments, and equal loading was confirmed by westernblotting for β-actin. Error bars represent standard error values from three different experiments. Scale bars=50 μm. ⁎pb0.05 compared to control (Student's t-test).

102 Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

(Fig. 3E) using an antibody specific to MT-I/-II isoforms confirmedthat MT treatment led to elevated levels of MT within astrocytes (weacknowledge that some of this may be the exogenously applied MT).

MT mediates astrogliosis through a JAK/STAT and RhoA dependentpathway

To screen for potential biochemical pathways mediating MT-inducedastrogliosis, we established astrocyte cultures in 96-well plates. Ratastrocytes were pre-treated a variety of specific inhibitors targetingMAPK (PD98059: 10 μM), PI3-kinase (LY29402: 10 μM) or JAK/STAT(AG490:10 μM)signallingpathways, followedby treatmentwith5 μg/mlMT. After 24 h, GFAP levels within astrocytes were determined byquantitative immunocytochemistry (Fig. 4A). MT-induced increases inGFAP expression was blocked by an inhibitor of JAK/STAT but not MAPKor PI3 kinase signalling (n=3). Subsequent experiments demonstratedthat the effect of the JAK/STAT inhibitor was dose dependent, and aconcentration of 20 μM could almost completely block MT-mediatedGFAP expression (Fig. 4B). Western blotting confirmed an increase inJAK1 levels and phosphorylated STAT3(tyr 705) within astrocytes 2 h aftertreatment with exogenous MT (Fig. 4C). Similarly, western blottingconfirmed that AG490 blockedMT-induced elevation inGFAP expression(results not shown).

Another key intracellular pathway associated with regulatingastroglial morphology is rho kinase. When MT was added to culturedrat cortical astrocytes, levels of GTP-RhoA were significantly reducedcompared to saline-treated astrocytes (Fig. 4D). The inhibitory effectof MT upon RhoA activation was mildly greater than that of IL-1β, aknown inhibitor of RhoA activity in astrocytes (Fig. 4D).

MT treatment alters astrocytic permissiveness to neurite outgrowth

To investigate the functional phenotype of MT-activated astro-cytes, confluent rat cortical astrocytes were treated with 5 μg/ml MT

for 24 h, followed by a media change and the seeding of 1×104

cortical neurons into the culture. After a further 24 h, neuriteoutgrowth of cortical neurons upon the astrocyte feeder layer wasassessed. Compared to neurons plated onto untreated astrocytes (Fig.5A), those plated onto astrocytes pre-treated with MT extendedlonger neurites within 24 h (Fig. 5B). As a further comparison,astrocytes were treated with the cytokine TGF-β1 (20 ng/ml), whichhas been demonstrated to induce astrogliosis (O'Toole et al., 2007)prior to seeding with cortical neurons. The TGF-β1 treatment resultedin astrocytes becoming less permissive to neurite outgrowth (Fig. 5C).Interestingly however, when TGF-β1-treated astrocytes were treatedwith MT for 24 h, these astrocytes becamemore permissive to neuriteoutgrowth (pb0.05; Fig. 5C).

The permissivity of astrocytes to provide a substrate for theattachment and growth of neurons is dependent upon the expressionof cell adhesion and growth inhibitory molecules, of which chondroi-tin-sulphate proteoglycans (CSPGs) represent the major familyexpressed by reactive astrocytes (reviewed Yiu and He, 2006). Wetherefore investigated whether the ability of MT treatment to makeastrocytes more permissive to neurite outgrowth is related to theexpression profile of CSPGs in MT-treated astrocytes. MT treatmentsignificantly decreased CSPG expression by 48 h as determined byquantitative immunocytochemistry (Fig. 5D). As a positive control, weused TGF-β1 (20 ng/ml), which increased CSPG expression after 48 h(as reported previously, O'Toole et al., 2007). In agreement with theneurite outgrowth studies (Fig. 5C), elevated CSPG expressioninduced by TGF-β1 treatment could be blocked by subsequenttreatment with MT for 24 h (Fig. 5D). To determine which specificinhibitory CSPG subtypes are affected byMT treatment, we performedwestern blotting of astrocyte lysates for neurocan and phosphacan.We found that TGF-β1 treatment induced expression of both of theseCSPG isoforms by astrocytes (Fig. 6), as has been reported previously(Asher et al., 2001; Smith and Strunz, 2005). Treatment with MTblocked TGF-β1 induced upregulation of both neurocan and

Fig. 4. Astrocytes were pre-treated with 10 μM of either PD98059, LY29402 or AG490,followed by addition of MT (A). AG490 significantly blocked MTmediated upregulationof GFAP by approximately 40% as measured by quantitative immunocytochemistry. In afurther series of experiments, it was found that AG490 dose-dependently blocked MT-mediated astrogliosis, with approximately 80% inhibition at a concentration of 20 μM(B). Western blotting (representative of three experiments) demonstrated that MTtreatment resulted in elevated levels of JAK1 and phosphorylated STAT3(tyr 705) (C). Inparallel experiments using a RhoA activation assay, it was found that MT treatmentsignificantly reduced levels of activated RhoA (D). Total RhoA levels (determined fromthe same samples that did not undergo the activated RhoA assay) were not altered (D).For all western blotting experiments, 10 μg total protein per sample was loaded. Errorbars represent standard error values from three different experiments. ⁎pb0.05compared to control (ANOVA with Tukey's post-hoc test). #pb0.05 between treatmentgroups.

Fig. 5. Confluent astrocytes were treated with either saline, 5 μg/ml MT, or 20 ng/mlTGF-β1 for 24 h, followed by a media change and subsequent plating of cortical neuronsinto the cultures at a density of 1×104 cells/well. After 24 h, neurite outgrowth wasassessed. MT treatment caused astrocytes to become more permissive to neuriteoutgrowth, assessed by immunocytochemistry (A, B; arrows indicate individualneurites) and by automated neurite outgrowth measurements performed using HCA-Vision (C). TGF-β1 treatment caused astrocytes to become less permissive to neuriteoutgrowth (C). When cytokine-treated astrocytes were treated with MT for 24 h,subsequent neurite outgrowth was significantly enhanced (C). To investigate how MTtreatment alters astrocytic permissivity to neurite outgrowth, confluent astrocyteswere treated with MT, and CSPG expression levels measured by quantitativeimmunocytochemistry 48 h after treatment. MT treatment caused a reduction inCSPG expression, which was significantly increased following TGF-β1 treatment (D).Interestingly, when TGF-β1 treatment was followed by subsequent treatment with MTthis resulted in decreased CSPG expression (D). Scale bars=50 μm (A, B). Error barsrepresent standard error values from three different experiments. ⁎pb0.05 comparedto control (ANOVA with Tukey's post-hoc test). #pb0.05 between treatment groups.

103Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

phosphacan in astrocytes, although the magnitude of MT effectappeared greatest upon phosphacan expression (Fig. 6). As a controlexperiment, we found that exposure of MT to non-cytokine treatedastrocytes had a minimal effect upon neurocan and phosphacanexpression (results not shown).

In a final series of experiments, we investigated whether thepresence of endogenousMT is able tomodulate astroglial permissivityto neurite outgrowth. Cultures of cortical astrocytes were obtainedfrom wildtype and MT null mutant (MTKO) mice and grown toconfluence, at which time embryonic cortical neurons from MTKOmice (to ensure that endogenous MT could only come from theunderlying wildtype astrocytes) were plated onto the astrocytes andneurite outgrowth assessed after 24 h. While the number of neuronsthat attached and extended neurites on the astroglial layer was notsignificantly different between wildtype and MTKO astrocytes (Fig.7A), the total neurite outgrowth was significantly reduced in neuronsplated onto MTKO astrocytes (Fig. 7B). Furthermore, the averagelength of neurites was significantly shorter in neurons plated onto theMTKO astrocytes (Fig. 7C). These results indicate that the endogenous

expression of MT by astrocytes alters astroglial permissivity to neuriteoutgrowth.

Discussion

The major findings of this study are that exogenous MT inducesmorphological changes and GFAP upregulation in cultured astrocytes

Fig. 6. Western blotting revealed that expression of the inhibitory CSPGs phosphacanand neurocanwas upregulated 48 h after treatment with 20 ng/ml TGF-β1. However, incultures where 5 μg/ml MT was applied to TGF-β1 treated astrocytes (24 h after initialtreatment with TGF-β1), phosphacan and neurocan expression was greatly reduced48 h after the initial treatment with TGF-β1. Western blotting for β-actin was used toensure that similar amounts of protein were loaded for each sample.

104 Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

that are consistent with astrogliosis. However, the functionalphenotype of MT-induced astrogliosis is permissive to neuriteoutgrowth, is associated with downregulation of CSPGs and isregulated via JAK/STAT and Rho signalling pathways. We thusdemonstrate that MT administration induces a pro-regenerativeastrogliosis. The MT-induced astrocytic response contrasts with the“classical” growth inhibitory astrogliosis that is associated with up-regulation of regeneration-inhibitory CSGPs and which is regulatedvia the MAPK signalling pathway.

Substantial evidence is accumulating in various CNS injury modelsto suggest a link between MT expression by astrocytes and positive,rather than negative, outcomes. Although MT expression is low in theuninjured brain, it is greatly elevated in reactive astrocytes bordering,and at some distance from, a cortical injury (Chung et al., 2004).Furthermore, while MT-overexpression confers neuroprotection aftertraumatic or ischaemic CNS injury (Giralt et al., 2002; Penkowa et al.,2002; van Lookeren Campagne et al., 1999), damage is exacerbated inMT knockouts (Penkowa et al., 1999; Natale et al., 2004; Carrasco etal., 2000; Trendelenburg et al., 2002). However, it was unclear howastrocytic MT, whichwas viewed in the past as an intracellular protein(Palmiter, 1998), could act to protect neuronswithin the injured brainand induce them to regenerate axons. In several recent studies, wehave demonstrated that extracellular MT secreted by reactiveastrocytes is capable of directly interacting with neurons via low-density lipoprotein receptors to promote axonal regeneration (Chunget al., 2003, 2008).

In the current study we have investigated whether exogenous MT,at levels comparable to those that might be secreted within the

Fig. 7. Embryonic cortical neurons were plated directly onto confluent cultures of astrocytesimmunocytochemical labelling of the axonal protein tau and automated neurite outgrowthsubstrate was not different betweenwildtype andMTKO astrocytes (A), the total amount of nthe average length of neurites was significantly reduced in neurons growing on MTKO astrotwo replicate experiments. ⁎pb0.05 compared to wildtype (t-test).

injured brain, is capable of altering astroglial responses to traumaticbrain injury. MT is present at 40 μg/g in normal human brain tissue(Erickson et al., 1994) suggesting that μg/ml concentrations in theextracellular space are conceivable after human brain injury (of note,MT levels are much lower in the rodent and murine brain). The finalavailable concentrations we used were estimated to be at physiolog-ical levels and our major in vitro observation was that exogenous MTinduced two hallmarks of astrogliosis, namely a striking hypertrophicmorphology and significant increases in GFAP expression.

We then showed by western blotting for activated JAK1 and STAT3that exogenous MT-induced astrogliotic changes were associatedwith activation of a JAK/STAT signalling pathway. Furthermore, wedemonstrated that the action of MT could be blocked using the JAK/STAT inhibitor AG490. These observations were at first surprising,since it is generally considered that signalling agents, such ascytokines which induce astrogliosis (notably astrogliosis that isinhibitory to axon growth), act via a MAPK-dependent pathway(Mandell et al., 2001). However, the JAK/STAT pathway has recentlybeen reported to be involved in regulating astrogliotic changes such aselevated GFAP expression in brain slice cultures (Damiani andO'Callaghan, 2007). Intriguingly, a recent study has reported thatconditional ablation of STAT3 following contusive spinal cord injurycauses limited migration of reactive astrocytes, which was associatedwith widespread infiltration of inflammatory cells, demyelination andsevere loss of motor function (Okada et al., 2006). Furthermore,conditional ablation of SOCS3, a negative feedback regulator of STAT3,prolonged STAT3 expression in reactive astrocytes and significantlyimproved wound healing and motor function. This suggests thatSTAT3 acts as a key regulator in reactive astrocytes to promote woundhealing rather than a detrimental response. In the context of ourstudy, this suggests that astrogliosis activated via a JAK/STATsignalling is permissive to regeneration and promotes recovery,while astrogliosis induced via a MAPK-dependent pathway isinhibitory to regeneration. Interestingly, it has recently been reportedthat exogenous MT treatment reduces STAT3 activation in culturedcerebellar granule neurons (Asmussen et al., 2009). In fact, MTappears to activate MAPK signalling pathways in neurons, via directinteraction with the low density lipoprotein receptor megalin (LRP2)and to a lesser extent the related receptor LRP1 (Ambjørn et al., 2008).Whether a similar low density lipoprotein receptor-mediated mech-anism is involved in our observation that MT induces astroglioticchanges is unclear and warrants further investigation, although it isknown that astrocytes do express both megalin and LRP1 (Bento-Abreu et al., 2008, 2009). Notably, there is a single report linkingligand binding to LRP and activation of the JAK/STAT pathway inmuscle and endothelial cells (Degryse et al., 2004). It is conceivablethen that MT signalling upon astrocytes might be mediated by LRP1,

from either wildtype or MT null (MTKO) astrocytes, and neurite outgrowth assessed bymeasurements (HCA-Vision). While the number of neurons present on the astroglialeurites was significantly reduced in neurons growing onMTKO astrocytes (B). Similarly,cytes (C). Data is from 12 replicate wells per condition. A similar result was observed in

105Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

and MT signalling upon neurons mediated primarily by interactionwith megalin, which would explain the activation of differentsignalling pathways by MT upon these neural cell types.

It is now well established that the Rho family of GTPases arecentral regulators of cell morphology. Since MT treatment was foundto induce pronounced astroglial hypertrophy, we also investigatedwhether MT alters Rho signalling within astrocytes. In accordancewith our immunocytochemical observations, MT significantly blockedRhoA activation but not total RhoA levels. Furthermore, this inhibitoryaction upon RhoA signalling was mildly greater than that imparted byIL-1β, which has previously been demonstrated to block RhoAsignalling and consequently alter astroglial morphology (John et al.,2004).

A key question that remained was what is the functionalphenotype of MT-activated astrocytes? This is critically importantsince astrogliosis and glial scar formation are generally considered asone of the fundamental inhibitory processes to neuroregeneration(Yiu and He, 2006), whereas our evidence suggested that astrogliosismight in fact be growth-permissive. We used a neurite outgrowthassay and showed that MT-treated astrocytes provided a significantlymore permissive substrate than vehicle- or cytokine-treated astro-cytes. Neurite outgrowth appeared to be mediated in part by areduced CSPG expression in response to MT treatment. Takentogether, the results suggest that MT is capable of inducing a pro-regenerative phenotype of reactive astrocytes and does so via theJAK/STAT signalling pathway. Intriguingly, we demonstrate thatMTKO astrocytes provide a less permissive substrate to neuriteoutgrowth than wildtype astrocytes. Since we have recently reportedthat cultured astrocytes secrete low levels of MT into culture medium(Chung et al., 2008), we think that this might explain the difference inpermissivity between MTKO and wildtype astrocytes that we haveobserved. In this scenario, astrocyte-secreted MT might signal uponastrocytes via the JAK/STAT pathway to make astrocytes morepermissive to neurite outgrowth, or secreted MT may act directlyupon the neurons themselves to promote neurite outgrowth (Chunget al., 2003, 2008; Ambjørn et al., 2008).

Interestingly, exogenous MT treatment induced endogenous MT-I/-II expression in astrocytes. We have recently reported that cytokinestimulation regulates secretion of endogenous MT-I/-II by astrocytes(Chung et al., 2008). The finding suggests that, under physiologicalconditions, MT might act in an autocrine or paracrine signallingmanner upon astrocytes. Based upon our observation that MT (eitherexogenous or endogenously expressed by astrocytes) makes astro-cytes more permissive to neurite outgrowth, we hypothesise that MTmay have a role in promoting regenerative sprouting in the injuredbrain. Indeed, this idea would fit with the observation that MTexpression by astrocytes following traumatic brain injury does notoccur immediately after the injury, but occurs 4 days later, concurrentwith the regenerative sprouting response that occurs post-injury(Chung et al., 2004). This hypothesis might also explain ourobservations after neural injury in vivo and in vitro whereby MTexpression increases in astrocytes first in the immediate vicinity of theinjury and then at increasing distances from it (Chung et al., 2004).The absence of such MT signalling may also partially explain whyMT-I/-II null mutant mice exhibit such long-term and wide-spreaddamage following traumatic brain injury with abnormalities in glialcell responses persisting for many weeks after the initial insult andextensive cell death spreading to the ipsilateral cortex (Penkowa etal., 1999).

In summary, evidence is accumulating to suggest that theastrocytic protein MT plays an important role in promoting recoverywithin the injured brain. We now show that one of these rolesinvolves MT directly inducing a type of reactive astrogliosis that isgrowth permissive rather than inhibitory. This finding may provideinsights into modulating astrogliosis therapeutically following braininjury.

Acknowledgments

This research was supported by the Australian Research Councilthrough a Linkage (LP0774820) project grant in partnership withBestenbalt LLC (Estonia), and Discovery (DP0556630, DP0984673)project grants. It was also supported by an Alzheimer's Australia grant(Jack & Ethel Goldin Foundation). SAD holds a National Health &Medical Research Council Senior Research Fellowship (ID: 254670).RSC holds a National Health &Medical Research Council Peter DohertyFellowship (ID: 352623).

References

Ambjørn, M., Asmussen, J.W., Lindstam, M., Gotfryd, K., Jacobsen, C., Kiselyov, V.V.,Moestrup, S.K., Penkowa, M., Bock, E., Berezin, V., 2008. Metallothionein and apeptide modeled after metallothionein, EmtinB, induce neuronal differentiationand survival through binding to receptors of the low-density lipoprotein receptorfamily. J. Neurochem. 104, 21–37.

Asher, R.A., Morgenstern, D.A., Moon, L.D., Fawcett, J.W., 2001. Chondroitin sulphateproteoglycans: inhibitory components of the glial scar. Prog. Brain Res. 132,611–619.

Asmussen, J.W., Von Sperling, M.L., Penkowa, M., 2009. Intraneuronal signalingpathways of metallothionein. J. Neurosci. Res. 87 (13), 2926–2936.

Bento-Abreu, A., Velasco, A., Polo-Hernández, E., Pérez-Reyes, P.L., Tabernero, A.,Medina, J.M., 2008. Megalin is a receptor for albumin in astrocytes and is requiredfor the synthesis of the neurotrophic factor oleic acid. J. Neurochem. 106,1149–1159.

Bento-Abreu, A., Velasco, A., Polo-Hernández, E., Lillo, C., Kozyraki, R., Tabernero, A.,Medina, J.M., 2009. Albumin endocytosis via megalin in astrocytes is caveola- andDab-1 dependent and is required for the synthesis of the neurotrophic factor oleicacid. J. Neurochem. 111 (1), 49–60.

Bush, T.G., Puvanachandra, N., Horner, C.H., Polito, A., Ostenfeld, T., Svendsen, C.N.,Mucke, L., Johnson, M.H., Sofroniew, M.V., 1999. Leukocyte infiltration, neuronaldegeneration, and neurite outgrowth after ablation of scar-forming, reactiveastrocytes in adult transgenic mice. Neuron 25, 297–308.

Carrasco, J., Penkowa, M., Hadberg, H., Molinero, A., Hidalgo, J., 2000. Enhanced seizuresand hippocampal neurodegeneration following kainic acid-induced seizures inmetallothionein-I + II-deficient mice. Eur. J. Neurosci. 12, 2311–2322.

Chung, R.S., Holloway, A.F., Eckhardt, B.L., Harris, J.A., Vickers, J.C., Chuah, M.I., West,A.K., 2002a. Sheep have an unusual variant of the brain-specific metallothionein,metallothionein-III. Biochem. J. 365, 323–328.

Chung, R.S., Vickers, J.C., Chuah, M.I., Eckhardt, B.L., West, A.K., 2002b. Metallothionein-III inhibits initial neurite formation in developing neurons as well as postinjury,regenerative neurite sprouting. Exp. Neurol. 178, 1–12.

Chung, R.S., Vickers, J.C., Chuah, M.I., West, A.K., 2003. Metallothionein-IIA promotesinitial neurite elongation and postinjury reactive neurite growth and facilitateshealing after focal cortical brain injury. J. Neurosci. 23, 3336–3342.

Chung, R.S., Adlard, P.A., Dittmann, J., Vickers, J.C., Chuah, M.I., West, A.K., 2004.Neuron–glia communication: metallothionein expression is specifically up-regulated by astrocytes in response to neuronal injury. J. Neurochem. 88,454–461.

Chung, R.S., Fung, S.J., Leung, Y.K., Walker, A.K., McCormack, G.H., Chuah, M.I., Vickers,J.C., West, A.K., 2007. Metallothionein expression by NG2 glial cells following CNSinjury. Cell. Mol. Life Sci. 64, 2716–2722.

Chung, R.S., Penkowa, M., Dittmann, J., King, C.E., Bartlett, C., Asmussen, J.W., Hidalgo,J., Carrasco, J., Leung, Y.K., Walker, A.K., Fung, S.J., Dunlop, S.A., Fitzgerald, M.,Beazley, L.D., Chuah, M.I., Vickers, J.C., West, A.K., 2008. Redefining the role ofmetallothionein within the injured brain: extracellular metallothioneins play animportant role in the astrocyte-neuron response to injury. J. Biol. Chem. 283,15349–15358.

Damiani, C.L., O'Callaghan, J.P., 2007. Recapitulation of cell signaling events associatedwith astrogliosis using the brain slice preparation. J. Neurochem. 100, 720–726.

Degryse, B., Neels, J.G., Czekay, R.P., Aertgeerts, K., Kamikubo, Y., Loskutoff, D.J., 2004.The low density lipoprotein receptor-related protein is a motogenic receptor forplasminogen activator inhibitor-1. J. Biol. Chem. 279, 22595–22604.

Erickson, J.C., Sewell, A.K., Jensen, L.T., Winge, D.R., Palmiter, R.D., 1994. Enhancedneurotrophic activity in Alzheimer's disease cortex is not associated with down-regulation of metallothionein-III (GIF). Brain Res. 649, 297–304.

Giralt, M., Penkowa, M., Lago, N., Molinero, A., Hidalgo, J., 2002. Metallothionein-1+2protect the CNS after a focal brain injury. Exp. Neurol. 173, 114–128.

John, G.R., Chen, L., Rivieccio, M.A., Melendez-Vasquez, C.V., Hartley, A., Brosnan, C.F.,2004. Interleukin-1beta induces a reactive astroglial phenotype via deactivation ofthe Rho GTPase-Rock axis. J. Neurosci. 24, 2837–2845.

Kohler, L.B., Berezin, V., Bock, E., Penkowa, M., 2003. The role of metallothionein II inneuronal differentiation and survival. Brain Res. 992, 128–136.

Mandell, J.W., Gocan, N.C., Vandenberg, S.R., 2001. Mechanical trauma induces rapidastroglial activation of ERK/MAP kinase: evidence for a paracrine signal. Glia 34,283–295.

Michalska, A.E., Choo, K.H., 1993. Targeting and germ-line transmission of a nullmutation at the metallothionein I and II loci in mouse. Proc. Natl. Acad. Sci. U. S. A.90, 8088–8092.

106 Y.K.J. Leung et al. / Experimental Neurology 221 (2010) 98–106

Natale, J.E., Knight, J.B., Cheng, Y., Rome, J.E., Gallo, V., 2004. Metallothionein I and IImitigate age-dependent secondary brain injury. J. Neurosci. Res. 78, 303–314.

Okada, S., Nakamura, M., Katoh, H., Miyao, T., Shimazaki, T., Ishii, K., Yamane, J.,Yoshimura, A., Iwamoto, Y., Toyama, Y., Okano, H., 2006. Conditional ablation ofStat3 or Socs3 discloses a dual role for reactive astrocytes after spinal cord injury.Nat. Med. 12, 829–834.

O'Toole, D.A., West, A.K., Chuah, M.I., 2007. Effect of olfactory ensheathing cells onreactive astrocytes in vitro. Cell. Mol. Life Sci. 64, 1303–1309.

Palmiter, R.D., 1998. The elusive function of metallothioneins. Proc. Natl. Acad. Sci.U. S. A. 95, 8428–8430.

Penkowa, M., Carrasco, J., Giralt, M., Moos, T., Hidalgo, J., 1999. CNS wound healing isseverely depressed in metallothionein I- and II-deficient mice. J. Neurosci. 19,2535–2545.

Penkowa, M., Giralt, M., Camats, J., Hidalgo, J., 2002. Metallothionein 1+2 protect theCNS during neuroglial degeneration induced by 6-aminonicotinamide. J. Comp.Neurol. 444, 174–189.

Penkowa, M., Florit, S., Giralt, M., Quintana, A., Molinero, A., Carrasco, J., Hidalgo, J., 2005.Metallothionein reduces central nervous system inflammation, neurodegenera-tion, and cell death following kainic acid-induced epileptic seizures. J. Neurosci.Res. 79, 522–534.

Penkowa, M., Cáceres, M., Borup, R., Nielsen, F.C., Poulsen, C.B., Quintana, A., Molinero,A., Carrasco, J., Florit, S., Giralt, M., Hidalgo, J., 2006. Novel roles for metallothionein-I + II (MT-I + II) in defense responses, neurogenesis, and tissue restoration aftertraumatic brain injury: insights from global gene expression profiling in wild-typeand MT-I + II knockout mice. J. Neurosci. Res. 84, 1452–1474.

Smith, G.M., Strunz, C., 2005. Growth factor and cytokine regulation of chondroitinsulfate proteoglycans by astrocytes. Glia 52, 209–218.

Trendelenburg, G., Prass, K., Priller, J., Kapinya, K., Polley, A., Muselmann, C., Ruscher,K., Kannbley, U., Schmitt, A.O., Castell, S., Wiegand, F., Meisel, A., Rosenthal, A.,Dirnagl, U., 2002. Serial analysis of gene expression identifies metallothionein-II asmajor neuroprotective gene in mouse focal cerebral ischemia. J. Neurosci. 22,5879–5888.

van Lookeren Campagne, M., Thibodeaux, H., van Bruggen, N., Cairns, B., Gerlai, R.,Palmer, J.T., Williams, S.P., Lowe, D.G., 1999. Evidence for a protective role ofmetallothionein-1 in focal cerebral ischemia. Proc. Natl. Acad. Sci. 96,12870–12875.

West, A.K., Chuah, M.I., Vickers, J.C., Chung, R.S., 2004. Protective role of metallothio-neins in the injured mammalian brain. Rev. Neurosci. 15, 157–166.

Yiu, G., He, Z., 2006. Glial inhibition of CNS axon regeneration. Nat. Rev., Neurosci. 7,617–627.